Comparison of absorbed dose extrapolation methods for mouse-to-human translation of radiolabelled macromolecules.

Background: Extrapolation of human absorbed doses (ADs) from biodistribution experiments on laboratory animals is used to predict the efficacy and toxicity profiles of new radiopharmaceuticals. Comparative studies between available animal-to-human dosimetry extrapolation methods are missing. We compared five computational methods for mice-to-human AD extrapolations, using two different radiopharmaceuticals, namely [In]CHX-DTPA-scFv78-Fc and [Ga]NODAGA-RGDyK. Human organ-specific time-integrated activity coefficients (TIACs) were derived from biodistribution studies previously conducted in our centre. The five computational methods adopted are based on simple direct application of mice TIACs to human organs (M1), relative mass scaling (M2), metabolic time scaling (M3), combined mass and time scaling (M4), and organ-specific allometric scaling (M5), respectively. For [Ga]NODAGA-RGDyK, these methods for mice-to-human extrapolations were tested against the ADs obtained on patients, previously published by our group. Lastly, an average [Ga]NODAGA-RGDyK-specific allometric parameter α was calculated from the organ-specific biological half-lives in mouse and humans and retrospectively applied to M3 and M4 to assess differences in human AD predictions with the α = 0.25 recommended by previous studies.

Results: For both radiopharmaceuticals, the five extrapolation methods showed significantly different AD results (p < 0.0001). In general, organ ADs obtained with M3 were higher than those obtained with the other methods. For [Ga]NODAGA-RGDyK, no significant differences were found between ADs calculated with M3 and those obtained directly on human subjects (H) (p = 0.99; average M3/H AD ratio = 1.03). All other methods for dose extrapolations resulted in ADs significantly different from those calculated directly on humans (all p ≤ 0.0001). Organ-specific allometric parameters calculated using combined experimental [Ga]NODAGA-RGDyK mice and human biodistribution data varied significantly. ADs calculated with M3 and M4 after the application of α = 0.17 were significantly different from those obtained by the application of α = 0.25 (both p < 0.001).

Conclusions: Available methods for mouse-to-human dosimetry extrapolations provided significantly different results in two different experimental models. For [Ga]NODAGA-RGDyK, the best approximation of human dosimetry was shown by M3, applying a metabolic scaling to the mouse organ TIACs. The accuracy of more refined extrapolation algorithms adopting model-specific metabolic scaling parameters should be further investigated.